メニュー
Conducts multi-round deep investigation of GitHub repositories, producing structured reports with timeline reconstruction, metrics analysis, and competitive comparisons.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | github-deep-research |
| description | Conducts multi-round deep investigation of GitHub repositories, producing structured reports with timeline reconstruction, metrics analysis, and competitive comparisons. |
| version | 1.0.0 |
| tags | ["research","github","code-analysis"] |
Multi-round research combining GitHub API, web_search, web_fetch to produce comprehensive markdown reports.
Broad to Narrow: Start with GitHub API, then general queries, refine based on findings.
Round 1: GitHub API
Round 2: "{topic} overview"
Round 3: "{topic} architecture", "{topic} vs alternatives"
Round 4: "{topic} issues", "{topic} roadmap", "site:github.com {topic}"
Source Prioritization:
Round 1 - GitHub API
Directly execute scripts/github_api.py without read_file():
python /path/to/skill/scripts/github_api.py <owner> <repo> summary
python /path/to/skill/scripts/github_api.py <owner> <repo> readme
python /path/to/skill/scripts/github_api.py <owner> <repo> tree
Available commands (the last argument of github_api.py):
Round 2 - Discovery (3-5 web_search)
Round 3 - Deep Investigation (5-10 web_search + web_fetch)
Round 4 - Deep Dive
Follow template in assets/report_template.md:
Include diagrams where helpful:
Timeline (Gantt):
gantt
title Project Timeline
dateFormat YYYY-MM-DD
section Phase 1
Development :2025-01-01, 2025-03-01
section Phase 2
Launch :2025-03-01, 2025-04-01
Architecture (Flowchart):
flowchart TD
A[User] --> B[Coordinator]
B --> C[Planner]
C --> D[Research Team]
D --> E[Reporter]
Comparison (Pie/Bar):
pie title Market Share
"Project A" : 45
"Project B" : 30
"Others" : 25
Assign confidence based on source quality:
| Confidence | Criteria |
|---|---|
| High (90%+) | Official docs, GitHub data, multiple corroborating sources |
| Medium (70-89%) | Single reliable source, recent articles |
| Low (50-69%) | Social media, unverified claims, outdated info |
Save report as: research_{topic}_{YYYYMMDD}.md
[citation:Title](URL) format immediately after each claim from external sourcesGood - With inline citations:
The project gained 10,000 stars within 3 months of launch [citation:GitHub Stats](https://github.com/owner/repo).
The architecture uses LangGraph for workflow orchestration [citation:LangGraph Docs](https://langchain.com/langgraph).
Bad - Without citations:
The project gained 10,000 stars within 3 months of launch.
The architecture uses LangGraph for workflow orchestration.
Reviews academic papers and preprints for methodology rigor, contribution significance, and reproducibility. Triggers on paper URLs, PDF uploads, arXiv links, or review requests.
Assesses the computational reproducibility of academic papers by detecting code repositories, building sandbox environments, running experiments, and comparing outputs against paper claims.
Analyzes research trends across years by searching academic databases, building trend curves, identifying emerging topics, and generating visualizations.
Conducts systematic literature reviews across multiple academic papers using PRISMA methodology, with structured synthesis and APA/IEEE/BibTeX citation support.
Guides autonomous decisions on when extracting and analyzing web images adds sufficient value to justify the token cost for visually dense research domains.
Creates, modifies, evaluates, and benchmarks agent skills including description optimization, eval testing, and performance variance analysis.